We further improve the final model’s agentic performance through parallel agents followed by self-verification.
1 more from this model
Parallel agent execution followed by self-verification improves Terminal-Bench 2.1 from 62.7% to 71.9%
specific method · filed under inference & serving
Parallel agents produce candidate work that is followed by self-verification.
The sources on this page only mention it in passing. The line above is the extraction's one-line gloss. See agentic scaffolding for the methods filed beside it.
2 quotes from 2 sources, one card per model, each opening with its strongest.
We further improve the final model’s agentic performance through parallel agents followed by self-verification.
Parallel agent execution followed by self-verification improves Terminal-Bench 2.1 from 62.7% to 71.9%
We further improve the final model’s agentic performance through parallel agents followed by self-verification.